KIT Career ServiceStudentsTheses

Noise Learning-Based Denoising for Approximate Message Passing

Research topic/area
Compressed Sensing, Signal Processing, Telecommunications
Type of thesis
Master
Start time
-
Application deadline
31.12.2027
Duration of the thesis
-

Description

Approximate Message Passing is a modern and powerful signal processing method that can recover signals from under-sampled measurements. An essential part of the AMP algorithm is the denoiser. The performance of AMP depends heavily on the choice of the denoiser, which must be selected according to the characteristics of the signal being recovered.

To accommodate complicated signals that cannot easily be defined in a closed form, deep-learning-based denoisers can be used. One such choice is autoencoders, which are already widely used for denoising. However, autoencoders tend to distort the images that they are used on. This makes them suboptimal denoisers for low-noise scenarios, which significantly affects the performance of the AMP.

Noise learning-based autoencoders overcome this flaw by learning the noise instead of the signal, therefore reducing the distortion in the high-SNR regions. In your master's thesis, you will investigate how noise learning-based autoencoders can be used to improve the performance of the AMP.

Tasks:
* Training autoencoders in Python,
* Finding ways to optimize the performance of these autoencoders to work as a denoiser with the AMP.
* Comparing the performance of the trained neural networks to existing works.

Requirement

Requirements for students
  • Foundations in probability theory and linear algebra,
  • Interest in approximate message passing and neural networks,
  • Interest in Python.

Faculty departments
  • Engineering sciences
    Electrical engineering & information technologies
    Electrical Engineering and Information Technology


Supervision

Title, first name, last name
Esen Özbay
Organizational unit
CEL
Email address
esen.oezbay@cel.kit.edu
Link to personal homepage/personal page
Website

Application via email

Application documents
  • Curriculum vitae
  • Grade transcript

E-Mail Address for application
Senden Sie die oben genannten Bewerbungsunterlagen bitte per Mail an esen.oezbay@cel.kit.edu


Back